Add llava support in ludwig

[skanjila]

Is your feature request related to a problem? Please describe. The feature request is to have multi modal capability inside ludwig using llama-next

Describe the use case The basic use cases are to add imagery/video on top of text and the llm workflow for training and inference

Describe the solution you'd like This PR will add the shell code to embed mulitmodal functionality into ludwig leveraging the current code thats already in there

Describe alternatives you've considered N/A

Additional context None